What Australia's $34 billion in reported cost overruns actually is

Over two decades, Australian governments have spent $34 billion more on transport infrastructure than they first told taxpayers they would. Across all projects worth $20 million or more, actual costs exceeded promised costs by 21 per cent. Among projects with an initial price tag above $1 billion, almost half overran — and those that did overran by 30 per cent on average.

Those numbers, from the Grattan Institute, are the ones everybody quotes. They are also almost useless for understanding what is actually going wrong, because they measure the distance between two things that are frequently not comparable: an announcement and an outcome.

A project that moves from $10 billion to $15 billion has not necessarily overrun by $5 billion. It may have added $3 billion of scope that government approved along the way, absorbed $1 billion of market escalation nobody could have priced in 2019, and overrun by $1 billion. Those are three different failures with three different fixes, and one of them is not a failure at all.

Nobody publishes the decomposition. That is the real problem.

What "cost increase" actually contains

Any movement between an announced figure and a final cost is some combination of four things:

1. Baseline failure. The original number was not an estimate. It was a political commitment made before there was enough design to estimate anything.

2. Approved scope growth. Government added scope and funded it. The project got bigger. This is a decision, not an overrun.

3. Market escalation. Labour, materials and plant cost more than forecast, often for reasons entirely outside the project.

4. Delivery overrun. The defined scope cost more to build than a sound estimate said it would.

Only the fourth is what most people mean by "cost overrun". It is also, on the evidence, probably the smallest of the four.

Root cause one: the baseline was never a real number

This is the largest single contributor, and the evidence is unusually direct.

Projects announced before government is prepared to formally commit are the problem. Grattan found that roughly one-third of projects were announced prematurely — and those projects account for more than three-quarters of all cost overruns.

Read that again. A third of the projects generate three-quarters of the problem, and the distinguishing feature is not what they were building or who was building it. It is that a number was put into the public domain before anyone could responsibly produce one.

The design maturity problem

For a cost estimate to carry meaningful confidence, design should typically be at 20 to 40 per cent maturity. In practice, costs have sometimes been estimated on design at 10 to 15 per cent maturity.

At 10 per cent design, you know roughly what you intend to build. You do not know the ground conditions, the services you will hit, the staging, the temporary works, or half of what the approvals will demand. An estimate at that maturity is a scoping exercise wearing the costume of a budget.

The estimate is then announced, quoted for years, and eventually used as the baseline against which "blowout" is calculated.

The contingency arithmetic does not work

This is the most damning finding in the whole area, and it is arithmetic rather than opinion.

Australian government projects seeking Commonwealth funding must present cost estimates as both a P50 (50 per cent confidence the final cost will not be exceeded) and a P90 (90 per cent confidence).

In business cases produced in recent years, the gap between P50 and P90 has generally been about 7 per cent.

Analysis of projects completed over the past two decades shows the actual difference between P50 and P90 costs averaged 49 per cent.

The contingency being carried is roughly one-seventh of what the historical record says it should be. A project funded at P50 with a 7 per cent band to P90 is not a project with modest risk. It is a project whose risk has been assessed with a model that does not match observed reality.

If the estimating framework systematically under-provisions by that margin, then "overruns" are not a delivery phenomenon. They are a forecasting phenomenon, showing up years later on someone else's watch.

The federal review said so explicitly

The 2023 Independent Strategic Review of the Infrastructure Investment Program found:

• $32.8 billion in known cost pressures across the program
• $14.2 billion of that on projects not yet under construction — many of which had not completed detailed cost estimates
• A ten-year pipeline that could not be delivered within its $120 billion allocation
• Many projects lacking "credible planning and costings"

Fifty projects were subsequently cancelled.

The detail worth holding onto: $14.2 billion of cost pressure on projects that had not started. Nothing had been built. No contractor had underperformed. The pressure was entirely a function of the numbers having been wrong when they were set.

Root cause two: approved scope growth counted as failure

Scope growth on major projects averages around 15 per cent, and the relationship is not linear — analysis suggests each 1 per cent of scope expansion drives about 1.6 per cent of cost growth.

That multiplier is the part practitioners recognise and commentators miss. Adding scope late does not cost what it would have cost if it had been in the original design. It costs more, because it disrupts a sequence that was already priced, procured and programmed.

Worked illustration. A project announced at $10 billion, delivered at $15 billion — a 50 per cent increase, and the headline writes itself.

Decompose it:

Same project. Same $5 billion. But "a 50 per cent blowout" and "an 11 per cent delivery overrun alongside a government decision to build something bigger" describe entirely different situations, and imply entirely different responses.

This example is illustrative, not an analysis of any specific project. That is precisely the point: the decomposition is almost never published, so nobody outside the project can do this arithmetic. The public gets one number, and that number carries the blame for four different things.

Root cause three: the market got more expensive, and the pipeline did it

Escalation is real, measurable, and largely outside any individual project's control.

RLB forecasts construction cost growth in 2026 of 4.0 per cent in Sydney and Melbourne, 5.0 per cent in Brisbane, 5.5 per cent on the Gold Coast, 5.3 per cent in Perth, 5.1 per cent in Adelaide and up to 6.0 per cent in Townsville.

Compounded across a six or eight-year build, that alone moves a project materially — and Snowy 2.0's original contract explicitly priced eight years of escalation and still doubled.

Infrastructure Australia's 2025 Market Capacity Report identifies what is driving it: an elevated near-term pipeline, skilled labour shortages, low productivity, insolvency risk, and limited competition among both Tier 1 contractors and subcontractors.

The most revealing figure in that report is about workforce:

Peak workforce demand has risen from 417,000 to 521,000 — and shifted out a full year, from mid-2026 to mid-2027. Infrastructure Australia's own reading is that this is "likely reflective of planned expenditure being pushed back as the market struggles to meet overly ambitious delivery targets."

That is the pipeline eating itself. Governments commit to more work than the market can deliver; the market cannot deliver it; the work slips; slipping work costs more; and the cost increase gets recorded against individual projects rather than against the decision to commit to all of them simultaneously.

Root cause four: how the work gets priced to win

This is where the industry's own conduct enters, and it deserves to be stated plainly rather than gestured at.

Contractors knowingly submit unrealistically low prices at tender to secure work, then use the design and construct process to recover margin through cost and time claims. The pattern is well documented, and it is not new — a spate of medium-to-large Australian contractor insolvencies between 2011 and 2013 is widely attributed in part to underpricing.

Two structural features sustain it:

Projects have grown past the point where a single Tier 1 can carry them. Rising contract values mean projects are increasingly too large for one contractor, which pushes work into consortia and joint ventures and reduces the number of parties capable of bidding at all. Fewer bidders is less competitive tension, and less competitive tension is higher prices — or the same prices with more risk priced out.

Risk is transferred down rather than managed. The risk profile a Tier 1 accepts upstream from government is generally more favourable than the profile it imposes downstream on subcontractors. The Tier 1 rarely absorbs the gap, because its downstream position covers its upstream exposure.

That is the subcontractor's whole problem in one sentence. Risk that government thought it had transferred to a contractor capable of managing it has frequently been passed to parties with neither the balance sheet nor the information to price it.

When those parties fail — and construction is persistently among Australia's worst sectors for insolvency — the cost does not disappear. It resurfaces as delay, retendered scope and claims, and lands back on the project as an "overrun".

What the trajectories look like

These are the published figures. They are cost increases, not overruns, and the distinction is the argument of this article.

In none of these cases is a public decomposition available showing how much is scope, how much is escalation, and how much is delivery performance against a sound baseline.

Some observations that can fairly be made:

The time between the early figure and the later one matters enormously. Inland Rail's $4.4 billion is a 2010 number and North East Link's $6 billion is a 2008 number. Comparing a 2008 estimate to a 2026 cost without adjusting for escalation is not analysis.

Cross River Rail moved fastest — $7.7 billion to $19 billion across roughly two years, with the largest revision arriving shortly after a change of government. Restating an inherited number is a different event from a cost increasing, and the public record does not distinguish them.

Snowy 2.0's contract explicitly priced escalation across an eight-year program at award, and still reset. That makes it the hardest of these to attribute to escalation alone.

What would actually fix it

The recommendations converge across the Grattan Institute, the Queensland Audit Office and the federal Strategic Review, which is itself informative.

Do not announce a number before there is a design to support it. One-third of projects announced prematurely producing three-quarters of overruns is about as clear as evidence gets.

Fund at a realistic confidence level. If the historical P50-to-P90 gap is 49 per cent and business cases are carrying 7, the estimating framework is the problem before any contractor is engaged.

Publish the decomposition. Governments should report cost movement split into approved scope change, escalation and delivery variance. It would be uncomfortable and it would end most of the sloppier reporting in a single stroke.

Governance with authority to test assumptions. The Queensland Audit Office is explicit: committees need power to test assumptions and act when cost, time, quality or scope deviate — and several entities lacked internal risk frameworks capable of detecting cost pressure early.

Package projects so more contractors can bid. Breaking megaprojects into elements that Tier 2 and Tier 3 contractors can competitively bid is being actively promoted, and it addresses the competition problem directly.

Stop transferring risk to parties who cannot price it. NSW procurement guidelines have begun shifting on construction risk allocation, and the industry position is increasingly that unpriceable risk transferred downstream returns as claims, disputes and insolvency.

What to watch

1. Whether any Australian government publishes a cost decomposition on a major project. It would be a first, and it would change the quality of every subsequent debate.
2. The mid-2027 workforce peak. Infrastructure Australia has already moved it once. If it moves again, the pipeline is still exceeding market capacity.
3. Whether project packaging actually widens the bidder field, or simply moves the same concentration down a tier.
4. Contingency levels in new business cases. If the P50-to-P90 gap is still around 7 per cent, nothing has been learned.

Published 22 September 2026 · Updated 23 September 2026